projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Pull define-node-cleanup into release branch
[linux-2.6-block.git] / fs / nfs / inode.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/nfs/inode.c
3 *
4 * Copyright (C) 1992 Rick Sladkey
5 *
6 * nfs inode and superblock handling functions
7 *
8 * Modularised by Alan Cox <Alan.Cox@linux.org>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
10 *
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
13 *
14 */
15
16#include <linux/config.h>
17#include <linux/module.h>
18#include <linux/init.h>
19
20#include <linux/time.h>
21#include <linux/kernel.h>
22#include <linux/mm.h>
23#include <linux/string.h>
24#include <linux/stat.h>
25#include <linux/errno.h>
26#include <linux/unistd.h>
27#include <linux/sunrpc/clnt.h>
28#include <linux/sunrpc/stats.h>
29#include <linux/nfs_fs.h>
30#include <linux/nfs_mount.h>
31#include <linux/nfs4_mount.h>
32#include <linux/lockd/bind.h>
33#include <linux/smp_lock.h>
34#include <linux/seq_file.h>
35#include <linux/mount.h>
36#include <linux/nfs_idmap.h>
37#include <linux/vfs.h>
38
39#include <asm/system.h>
40#include <asm/uaccess.h>
41
4ce79717 42#include "nfs4_fs.h"
1da177e4
LT		 43#include "delegation.h"
44
45#define NFSDBG_FACILITY NFSDBG_VFS
46#define NFS_PARANOIA 1
47
48/* Maximum number of readahead requests
49 * FIXME: this should really be a sysctl so that users may tune it to suit
50 * their needs. People that do NFS over a slow network, might for
51 * instance want to reduce it to something closer to 1 for improved
52 * interactive response.
53 */
54#define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
55
56static void nfs_invalidate_inode(struct inode *);
57static int nfs_update_inode(struct inode *, struct nfs_fattr *, unsigned long);
58
59static struct inode *nfs_alloc_inode(struct super_block *sb);
60static void nfs_destroy_inode(struct inode *);
61static int nfs_write_inode(struct inode *,int);
62static void nfs_delete_inode(struct inode *);
63static void nfs_clear_inode(struct inode *);
64static void nfs_umount_begin(struct super_block *);
65static int nfs_statfs(struct super_block *, struct kstatfs *);
66static int nfs_show_options(struct seq_file *, struct vfsmount *);
ada70d94 67static void nfs_zap_acl_cache(struct inode *);
1da177e4
LT		 68
69static struct rpc_program nfs_program;
70
71static struct super_operations nfs_sops = {
72 .alloc_inode = nfs_alloc_inode,
73 .destroy_inode = nfs_destroy_inode,
74 .write_inode = nfs_write_inode,
75 .delete_inode = nfs_delete_inode,
76 .statfs = nfs_statfs,
77 .clear_inode = nfs_clear_inode,
78 .umount_begin = nfs_umount_begin,
79 .show_options = nfs_show_options,
80};
81
82/*
83 * RPC cruft for NFS
84 */
85static struct rpc_stat nfs_rpcstat = {
86 .program = &nfs_program
87};
88static struct rpc_version * nfs_version[] = {
89 NULL,
90 NULL,
91 &nfs_version2,
92#if defined(CONFIG_NFS_V3)
93 &nfs_version3,
94#elif defined(CONFIG_NFS_V4)
95 NULL,
96#endif
97#if defined(CONFIG_NFS_V4)
98 &nfs_version4,
99#endif
100};
101
102static struct rpc_program nfs_program = {
103 .name = "nfs",
104 .number = NFS_PROGRAM,
105 .nrvers = sizeof(nfs_version) / sizeof(nfs_version[0]),
106 .version = nfs_version,
107 .stats = &nfs_rpcstat,
108 .pipe_dir_name = "/nfs",
109};
110
b7fa0554
AG		 111#ifdef CONFIG_NFS_V3_ACL
112static struct rpc_stat nfsacl_rpcstat = { &nfsacl_program };
113static struct rpc_version * nfsacl_version[] = {
114 [3] = &nfsacl_version3,
115};
116
117struct rpc_program nfsacl_program = {
118 .name = "nfsacl",
119 .number = NFS_ACL_PROGRAM,
120 .nrvers = sizeof(nfsacl_version) / sizeof(nfsacl_version[0]),
121 .version = nfsacl_version,
122 .stats = &nfsacl_rpcstat,
123};
124#endif /* CONFIG_NFS_V3_ACL */
125
1da177e4
LT		 126static inline unsigned long
127nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
128{
129 return nfs_fileid_to_ino_t(fattr->fileid);
130}
131
132static int
133nfs_write_inode(struct inode *inode, int sync)
134{
135 int flags = sync ? FLUSH_WAIT : 0;
136 int ret;
137
3da28eb1 138 ret = nfs_commit_inode(inode, flags);
1da177e4
LT		 139 if (ret < 0)
140 return ret;
141 return 0;
142}
143
144static void
145nfs_delete_inode(struct inode * inode)
146{
147 dprintk("NFS: delete_inode(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
148
fef26658
MF		 149 truncate_inode_pages(&inode->i_data, 0);
150
1da177e4
LT		 151 nfs_wb_all(inode);
152 /*
153 * The following should never happen...
154 */
155 if (nfs_have_writebacks(inode)) {
156 printk(KERN_ERR "nfs_delete_inode: inode %ld has pending RPC requests\n", inode->i_ino);
157 }
158
159 clear_inode(inode);
160}
161
1da177e4
LT		 162static void
163nfs_clear_inode(struct inode *inode)
164{
165 struct nfs_inode *nfsi = NFS_I(inode);
166 struct rpc_cred *cred;
167
168 nfs_wb_all(inode);
169 BUG_ON (!list_empty(&nfsi->open_files));
ada70d94 170 nfs_zap_acl_cache(inode);
1da177e4
LT		 171 cred = nfsi->cache_access.cred;
172 if (cred)
173 put_rpccred(cred);
174 BUG_ON(atomic_read(&nfsi->data_updates) != 0);
175}
176
177void
178nfs_umount_begin(struct super_block *sb)
179{
6a19275a 180 struct rpc_clnt *rpc = NFS_SB(sb)->client;
1da177e4
LT		 181
182 /* -EIO all pending I/O */
6a19275a 183 if (!IS_ERR(rpc))
1da177e4 184 rpc_killall_tasks(rpc);
b7fa0554
AG		 185 rpc = NFS_SB(sb)->client_acl;
186 if (!IS_ERR(rpc))
187 rpc_killall_tasks(rpc);
1da177e4
LT		 188}
189
190
191static inline unsigned long
192nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
193{
194 /* make sure blocksize is a power of two */
195 if ((bsize & (bsize - 1)) || nrbitsp) {
196 unsigned char nrbits;
197
198 for (nrbits = 31; nrbits && !(bsize & (1 << nrbits)); nrbits--)
199 ;
200 bsize = 1 << nrbits;
201 if (nrbitsp)
202 *nrbitsp = nrbits;
203 }
204
205 return bsize;
206}
207
208/*
209 * Calculate the number of 512byte blocks used.
210 */
211static inline unsigned long
212nfs_calc_block_size(u64 tsize)
213{
214 loff_t used = (tsize + 511) >> 9;
215 return (used > ULONG_MAX) ? ULONG_MAX : used;
216}
217
218/*
219 * Compute and set NFS server blocksize
220 */
221static inline unsigned long
222nfs_block_size(unsigned long bsize, unsigned char *nrbitsp)
223{
224 if (bsize < 1024)
225 bsize = NFS_DEF_FILE_IO_BUFFER_SIZE;
226 else if (bsize >= NFS_MAX_FILE_IO_BUFFER_SIZE)
227 bsize = NFS_MAX_FILE_IO_BUFFER_SIZE;
228
229 return nfs_block_bits(bsize, nrbitsp);
230}
231
232/*
233 * Obtain the root inode of the file system.
234 */
235static struct inode *
236nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
237{
238 struct nfs_server *server = NFS_SB(sb);
239 struct inode *rooti;
240 int error;
241
242 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
243 if (error < 0) {
244 dprintk("nfs_get_root: getattr error = %d\n", -error);
245 return ERR_PTR(error);
246 }
247
248 rooti = nfs_fhget(sb, rootfh, fsinfo->fattr);
249 if (!rooti)
250 return ERR_PTR(-ENOMEM);
251 return rooti;
252}
253
254/*
255 * Do NFS version-independent mount processing, and sanity checking
256 */
257static int
258nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
259{
260 struct nfs_server *server;
261 struct inode *root_inode;
262 struct nfs_fattr fattr;
263 struct nfs_fsinfo fsinfo = {
264 .fattr = &fattr,
265 };
266 struct nfs_pathconf pathinfo = {
267 .fattr = &fattr,
268 };
269 int no_root_error = 0;
270 unsigned long max_rpc_payload;
271
272 /* We probably want something more informative here */
273 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
274
275 server = NFS_SB(sb);
276
277 sb->s_magic = NFS_SUPER_MAGIC;
278
279 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
280 /* Did getting the root inode fail? */
281 if (IS_ERR(root_inode)) {
282 no_root_error = PTR_ERR(root_inode);
283 goto out_no_root;
284 }
285 sb->s_root = d_alloc_root(root_inode);
286 if (!sb->s_root) {
287 no_root_error = -ENOMEM;
288 goto out_no_root;
289 }
290 sb->s_root->d_op = server->rpc_ops->dentry_ops;
291
292 /* Get some general file system info */
293 if (server->namelen == 0 &&
294 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
295 server->namelen = pathinfo.max_namelen;
296 /* Work out a lot of parameters */
297 if (server->rsize == 0)
298 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
299 if (server->wsize == 0)
300 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
301
302 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
303 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
304 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
305 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
306
307 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
308 if (server->rsize > max_rpc_payload)
309 server->rsize = max_rpc_payload;
310 if (server->wsize > max_rpc_payload)
311 server->wsize = max_rpc_payload;
312
313 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
314 if (server->rpages > NFS_READ_MAXIOV) {
315 server->rpages = NFS_READ_MAXIOV;
316 server->rsize = server->rpages << PAGE_CACHE_SHIFT;
317 }
318
319 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
320 if (server->wpages > NFS_WRITE_MAXIOV) {
321 server->wpages = NFS_WRITE_MAXIOV;
322 server->wsize = server->wpages << PAGE_CACHE_SHIFT;
323 }
324
325 if (sb->s_blocksize == 0)
326 sb->s_blocksize = nfs_block_bits(server->wsize,
327 &sb->s_blocksize_bits);
328 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
329
330 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
331 if (server->dtsize > PAGE_CACHE_SIZE)
332 server->dtsize = PAGE_CACHE_SIZE;
333 if (server->dtsize > server->rsize)
334 server->dtsize = server->rsize;
335
336 if (server->flags & NFS_MOUNT_NOAC) {
337 server->acregmin = server->acregmax = 0;
338 server->acdirmin = server->acdirmax = 0;
339 sb->s_flags |= MS_SYNCHRONOUS;
340 }
341 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
342
343 sb->s_maxbytes = fsinfo.maxfilesize;
344 if (sb->s_maxbytes > MAX_LFS_FILESIZE)
345 sb->s_maxbytes = MAX_LFS_FILESIZE;
346
347 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
348 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
349
350 /* We're airborne Set socket buffersize */
351 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
352 return 0;
353 /* Yargs. It didn't work out. */
354out_no_root:
355 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
356 if (!IS_ERR(root_inode))
357 iput(root_inode);
358 return no_root_error;
359}
360
361/*
362 * Create an RPC client handle.
363 */
364static struct rpc_clnt *
365nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
366{
367 struct rpc_timeout timeparms;
368 struct rpc_xprt *xprt = NULL;
369 struct rpc_clnt *clnt = NULL;
370 int tcp = (data->flags & NFS_MOUNT_TCP);
371
372 /* Initialize timeout values */
373 timeparms.to_initval = data->timeo * HZ / 10;
374 timeparms.to_retries = data->retrans;
375 timeparms.to_maxval = tcp ? RPC_MAX_TCP_TIMEOUT : RPC_MAX_UDP_TIMEOUT;
376 timeparms.to_exponential = 1;
377
378 if (!timeparms.to_initval)
379 timeparms.to_initval = (tcp ? 600 : 11) * HZ / 10;
380 if (!timeparms.to_retries)
381 timeparms.to_retries = 5;
382
383 /* create transport and client */
384 xprt = xprt_create_proto(tcp ? IPPROTO_TCP : IPPROTO_UDP,
385 &server->addr, &timeparms);
386 if (IS_ERR(xprt)) {
9085bbcb
TM		 387 dprintk("%s: cannot create RPC transport. Error = %ld\n",
388 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 389 return (struct rpc_clnt *)xprt;
390 }
391 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
392 server->rpc_ops->version, data->pseudoflavor);
393 if (IS_ERR(clnt)) {
9085bbcb
TM		 394 dprintk("%s: cannot create RPC client. Error = %ld\n",
395 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 396 goto out_fail;
397 }
398
399 clnt->cl_intr = 1;
400 clnt->cl_softrtry = 1;
401 clnt->cl_chatty = 1;
402
403 return clnt;
404
405out_fail:
1da177e4
LT		 406 return clnt;
407}
408
409/*
410 * The way this works is that the mount process passes a structure
411 * in the data argument which contains the server's IP address
412 * and the root file handle obtained from the server's mount
413 * daemon. We stash these away in the private superblock fields.
414 */
415static int
416nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
417{
418 struct nfs_server *server;
419 rpc_authflavor_t authflavor;
420
421 server = NFS_SB(sb);
422 sb->s_blocksize_bits = 0;
423 sb->s_blocksize = 0;
424 if (data->bsize)
425 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
426 if (data->rsize)
427 server->rsize = nfs_block_size(data->rsize, NULL);
428 if (data->wsize)
429 server->wsize = nfs_block_size(data->wsize, NULL);
430 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
431
432 server->acregmin = data->acregmin*HZ;
433 server->acregmax = data->acregmax*HZ;
434 server->acdirmin = data->acdirmin*HZ;
435 server->acdirmax = data->acdirmax*HZ;
436
437 /* Start lockd here, before we might error out */
438 if (!(server->flags & NFS_MOUNT_NONLM))
439 lockd_up();
440
441 server->namelen = data->namlen;
442 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
443 if (!server->hostname)
444 return -ENOMEM;
445 strcpy(server->hostname, data->hostname);
446
447 /* Check NFS protocol revision and initialize RPC op vector
448 * and file handle pool. */
1da177e4 449#ifdef CONFIG_NFS_V3
9085bbcb 450 if (server->flags & NFS_MOUNT_VER3) {
1da177e4
LT		 451 server->rpc_ops = &nfs_v3_clientops;
452 server->caps |= NFS_CAP_READDIRPLUS;
1da177e4
LT		 453 } else {
454 server->rpc_ops = &nfs_v2_clientops;
455 }
9085bbcb
TM		 456#else
457 server->rpc_ops = &nfs_v2_clientops;
458#endif
1da177e4
LT		 459
460 /* Fill in pseudoflavor for mount version < 5 */
461 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
462 data->pseudoflavor = RPC_AUTH_UNIX;
463 authflavor = data->pseudoflavor; /* save for sb_init() */
464 /* XXX maybe we want to add a server->pseudoflavor field */
465
466 /* Create RPC client handles */
467 server->client = nfs_create_client(server, data);
468 if (IS_ERR(server->client))
469 return PTR_ERR(server->client);
470 /* RFC 2623, sec 2.3.2 */
471 if (authflavor != RPC_AUTH_UNIX) {
6a19275a
BF		 472 struct rpc_auth *auth;
473
1da177e4
LT		 474 server->client_sys = rpc_clone_client(server->client);
475 if (IS_ERR(server->client_sys))
476 return PTR_ERR(server->client_sys);
6a19275a
BF		 477 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
478 if (IS_ERR(auth))
479 return PTR_ERR(auth);
1da177e4
LT		 480 } else {
481 atomic_inc(&server->client->cl_count);
482 server->client_sys = server->client;
483 }
1da177e4 484 if (server->flags & NFS_MOUNT_VER3) {
b7fa0554
AG		 485#ifdef CONFIG_NFS_V3_ACL
486 if (!(server->flags & NFS_MOUNT_NOACL)) {
487 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
488 /* No errors! Assume that Sun nfsacls are supported */
489 if (!IS_ERR(server->client_acl))
490 server->caps |= NFS_CAP_ACLS;
491 }
492#else
493 server->flags &= ~NFS_MOUNT_NOACL;
494#endif /* CONFIG_NFS_V3_ACL */
055ffbea
AG		 495 /*
496 * The VFS shouldn't apply the umask to mode bits. We will
497 * do so ourselves when necessary.
498 */
499 sb->s_flags |= MS_POSIXACL;
1da177e4
LT		 500 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
501 server->namelen = NFS3_MAXNAMLEN;
502 sb->s_time_gran = 1;
503 } else {
504 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
505 server->namelen = NFS2_MAXNAMLEN;
506 }
507
508 sb->s_op = &nfs_sops;
509 return nfs_sb_init(sb, authflavor);
510}
511
512static int
513nfs_statfs(struct super_block *sb, struct kstatfs *buf)
514{
515 struct nfs_server *server = NFS_SB(sb);
516 unsigned char blockbits;
517 unsigned long blockres;
518 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
519 struct nfs_fattr fattr;
520 struct nfs_fsstat res = {
521 .fattr = &fattr,
522 };
523 int error;
524
525 lock_kernel();
526
527 error = server->rpc_ops->statfs(server, rootfh, &res);
528 buf->f_type = NFS_SUPER_MAGIC;
529 if (error < 0)
530 goto out_err;
531
532 /*
533 * Current versions of glibc do not correctly handle the
534 * case where f_frsize != f_bsize. Eventually we want to
535 * report the value of wtmult in this field.
536 */
537 buf->f_frsize = sb->s_blocksize;
538
539 /*
540 * On most *nix systems, f_blocks, f_bfree, and f_bavail
541 * are reported in units of f_frsize. Linux hasn't had
542 * an f_frsize field in its statfs struct until recently,
543 * thus historically Linux's sys_statfs reports these
544 * fields in units of f_bsize.
545 */
546 buf->f_bsize = sb->s_blocksize;
547 blockbits = sb->s_blocksize_bits;
548 blockres = (1 << blockbits) - 1;
549 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
550 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
551 buf->f_bavail = (res.abytes + blockres) >> blockbits;
552
553 buf->f_files = res.tfiles;
554 buf->f_ffree = res.afiles;
555
556 buf->f_namelen = server->namelen;
557 out:
558 unlock_kernel();
559
560 return 0;
561
562 out_err:
563 printk(KERN_WARNING "nfs_statfs: statfs error = %d\n", -error);
564 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
565 goto out;
566
567}
568
569static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
570{
571 static struct proc_nfs_info {
572 int flag;
573 char *str;
574 char *nostr;
575 } nfs_info[] = {
576 { NFS_MOUNT_SOFT, ",soft", ",hard" },
577 { NFS_MOUNT_INTR, ",intr", "" },
578 { NFS_MOUNT_POSIX, ",posix", "" },
579 { NFS_MOUNT_TCP, ",tcp", ",udp" },
580 { NFS_MOUNT_NOCTO, ",nocto", "" },
581 { NFS_MOUNT_NOAC, ",noac", "" },
582 { NFS_MOUNT_NONLM, ",nolock", ",lock" },
b7fa0554 583 { NFS_MOUNT_NOACL, ",noacl", "" },
1da177e4
LT		 584 { 0, NULL, NULL }
585 };
586 struct proc_nfs_info *nfs_infop;
587 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
588
589 seq_printf(m, ",v%d", nfss->rpc_ops->version);
590 seq_printf(m, ",rsize=%d", nfss->rsize);
591 seq_printf(m, ",wsize=%d", nfss->wsize);
592 if (nfss->acregmin != 3*HZ)
593 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
594 if (nfss->acregmax != 60*HZ)
595 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
596 if (nfss->acdirmin != 30*HZ)
597 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
598 if (nfss->acdirmax != 60*HZ)
599 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
600 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
601 if (nfss->flags & nfs_infop->flag)
602 seq_puts(m, nfs_infop->str);
603 else
604 seq_puts(m, nfs_infop->nostr);
605 }
606 seq_puts(m, ",addr=");
607 seq_escape(m, nfss->hostname, " \t\n\\");
608 return 0;
609}
610
611/*
612 * Invalidate the local caches
613 */
614void
615nfs_zap_caches(struct inode *inode)
616{
617 struct nfs_inode *nfsi = NFS_I(inode);
618 int mode = inode->i_mode;
619
dc59250c
CL		 620 spin_lock(&inode->i_lock);
621
1da177e4
LT		 622 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
623 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
624
625 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
626 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
55296809 627 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
1da177e4 628 else
55296809 629 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
dc59250c
CL		 630
631 spin_unlock(&inode->i_lock);
ada70d94
TM		 632}
633
634static void nfs_zap_acl_cache(struct inode *inode)
635{
636 void (*clear_acl_cache)(struct inode *);
637
638 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
639 if (clear_acl_cache != NULL)
640 clear_acl_cache(inode);
dc59250c 641 spin_lock(&inode->i_lock);
55296809 642 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
dc59250c 643 spin_unlock(&inode->i_lock);
1da177e4
LT		 644}
645
646/*
647 * Invalidate, but do not unhash, the inode
648 */
649static void
650nfs_invalidate_inode(struct inode *inode)
651{
652 umode_t save_mode = inode->i_mode;
653
654 make_bad_inode(inode);
655 inode->i_mode = save_mode;
656 nfs_zap_caches(inode);
657}
658
659struct nfs_find_desc {
660 struct nfs_fh *fh;
661 struct nfs_fattr *fattr;
662};
663
664/*
665 * In NFSv3 we can have 64bit inode numbers. In order to support
666 * this, and re-exported directories (also seen in NFSv2)
667 * we are forced to allow 2 different inodes to have the same
668 * i_ino.
669 */
670static int
671nfs_find_actor(struct inode *inode, void *opaque)
672{
673 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
674 struct nfs_fh *fh = desc->fh;
675 struct nfs_fattr *fattr = desc->fattr;
676
677 if (NFS_FILEID(inode) != fattr->fileid)
678 return 0;
679 if (nfs_compare_fh(NFS_FH(inode), fh))
680 return 0;
681 if (is_bad_inode(inode) || NFS_STALE(inode))
682 return 0;
683 return 1;
684}
685
686static int
687nfs_init_locked(struct inode *inode, void *opaque)
688{
689 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
690 struct nfs_fattr *fattr = desc->fattr;
691
692 NFS_FILEID(inode) = fattr->fileid;
693 nfs_copy_fh(NFS_FH(inode), desc->fh);
694 return 0;
695}
696
697/* Don't use READDIRPLUS on directories that we believe are too large */
698#define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
699
700/*
701 * This is our front-end to iget that looks up inodes by file handle
702 * instead of inode number.
703 */
704struct inode *
705nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
706{
707 struct nfs_find_desc desc = {
708 .fh = fh,
709 .fattr = fattr
710 };
711 struct inode *inode = NULL;
712 unsigned long hash;
713
714 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
715 goto out_no_inode;
716
717 if (!fattr->nlink) {
718 printk("NFS: Buggy server - nlink == 0!\n");
719 goto out_no_inode;
720 }
721
722 hash = nfs_fattr_to_ino_t(fattr);
723
724 if (!(inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc)))
725 goto out_no_inode;
726
727 if (inode->i_state & I_NEW) {
728 struct nfs_inode *nfsi = NFS_I(inode);
729
730 /* We set i_ino for the few things that still rely on it,
731 * such as stat(2) */
732 inode->i_ino = hash;
733
734 /* We can't support update_atime(), since the server will reset it */
735 inode->i_flags |= S_NOATIME|S_NOCMTIME;
736 inode->i_mode = fattr->mode;
737 /* Why so? Because we want revalidate for devices/FIFOs, and
738 * that's precisely what we have in nfs_file_inode_operations.
739 */
92cfc62c 740 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
1da177e4
LT		 741 if (S_ISREG(inode->i_mode)) {
742 inode->i_fop = &nfs_file_operations;
743 inode->i_data.a_ops = &nfs_file_aops;
744 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
745 } else if (S_ISDIR(inode->i_mode)) {
746 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
747 inode->i_fop = &nfs_dir_operations;
748 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
749 && fattr->size <= NFS_LIMIT_READDIRPLUS)
412d582e 750 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
1da177e4
LT		 751 } else if (S_ISLNK(inode->i_mode))
752 inode->i_op = &nfs_symlink_inode_operations;
753 else
754 init_special_inode(inode, inode->i_mode, fattr->rdev);
755
756 nfsi->read_cache_jiffies = fattr->timestamp;
757 inode->i_atime = fattr->atime;
758 inode->i_mtime = fattr->mtime;
759 inode->i_ctime = fattr->ctime;
760 if (fattr->valid & NFS_ATTR_FATTR_V4)
761 nfsi->change_attr = fattr->change_attr;
762 inode->i_size = nfs_size_to_loff_t(fattr->size);
763 inode->i_nlink = fattr->nlink;
764 inode->i_uid = fattr->uid;
765 inode->i_gid = fattr->gid;
766 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
767 /*
768 * report the blocks in 512byte units
769 */
770 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
771 inode->i_blksize = inode->i_sb->s_blocksize;
772 } else {
773 inode->i_blocks = fattr->du.nfs2.blocks;
774 inode->i_blksize = fattr->du.nfs2.blocksize;
775 }
776 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
777 nfsi->attrtimeo_timestamp = jiffies;
778 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
779 nfsi->cache_access.cred = NULL;
780
781 unlock_new_inode(inode);
782 } else
783 nfs_refresh_inode(inode, fattr);
784 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
785 inode->i_sb->s_id,
786 (long long)NFS_FILEID(inode),
787 atomic_read(&inode->i_count));
788
789out:
790 return inode;
791
792out_no_inode:
793 printk("nfs_fhget: iget failed\n");
794 goto out;
795}
796
797#define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
798
799int
800nfs_setattr(struct dentry *dentry, struct iattr *attr)
801{
802 struct inode *inode = dentry->d_inode;
803 struct nfs_fattr fattr;
804 int error;
805
806 if (attr->ia_valid & ATTR_SIZE) {
807 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
808 attr->ia_valid &= ~ATTR_SIZE;
809 }
810
811 /* Optimization: if the end result is no change, don't RPC */
812 attr->ia_valid &= NFS_VALID_ATTRS;
813 if (attr->ia_valid == 0)
814 return 0;
815
816 lock_kernel();
817 nfs_begin_data_update(inode);
818 /* Write all dirty data if we're changing file permissions or size */
819 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE)) != 0) {
820 if (filemap_fdatawrite(inode->i_mapping) == 0)
821 filemap_fdatawait(inode->i_mapping);
822 nfs_wb_all(inode);
823 }
824 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
65e4308d 825 if (error == 0)
1da177e4 826 nfs_refresh_inode(inode, &fattr);
65e4308d
TM		 827 nfs_end_data_update(inode);
828 unlock_kernel();
829 return error;
830}
831
832/**
833 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
834 * @inode: pointer to struct inode
835 * @attr: pointer to struct iattr
836 *
837 * Note: we do this in the *proc.c in order to ensure that
838 * it works for things like exclusive creates too.
839 */
840void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
841{
842 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
1da177e4 843 if ((attr->ia_valid & ATTR_MODE) != 0) {
65e4308d
TM		 844 int mode = attr->ia_mode & S_IALLUGO;
845 mode |= inode->i_mode & ~S_IALLUGO;
1da177e4
LT		 846 inode->i_mode = mode;
847 }
848 if ((attr->ia_valid & ATTR_UID) != 0)
849 inode->i_uid = attr->ia_uid;
850 if ((attr->ia_valid & ATTR_GID) != 0)
851 inode->i_gid = attr->ia_gid;
dc59250c 852 spin_lock(&inode->i_lock);
55296809 853 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
dc59250c 854 spin_unlock(&inode->i_lock);
65e4308d
TM		 855 }
856 if ((attr->ia_valid & ATTR_SIZE) != 0) {
857 inode->i_size = attr->ia_size;
858 vmtruncate(inode, attr->ia_size);
859 }
1da177e4
LT		 860}
861
412d582e
CL		 862static int nfs_wait_schedule(void *word)
863{
864 if (signal_pending(current))
865 return -ERESTARTSYS;
866 schedule();
867 return 0;
868}
869
1da177e4
LT		 870/*
871 * Wait for the inode to get unlocked.
1da177e4 872 */
412d582e 873static int nfs_wait_on_inode(struct inode *inode)
1da177e4
LT		 874{
875 struct rpc_clnt *clnt = NFS_CLIENT(inode);
876 struct nfs_inode *nfsi = NFS_I(inode);
412d582e 877 sigset_t oldmask;
1da177e4 878 int error;
412d582e 879
412d582e
CL		 880 rpc_clnt_sigmask(clnt, &oldmask);
881 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING,
882 nfs_wait_schedule, TASK_INTERRUPTIBLE);
883 rpc_clnt_sigunmask(clnt, &oldmask);
412d582e 884
1da177e4
LT		 885 return error;
886}
887
412d582e
CL		 888static void nfs_wake_up_inode(struct inode *inode)
889{
890 struct nfs_inode *nfsi = NFS_I(inode);
891
892 clear_bit(NFS_INO_REVALIDATING, &nfsi->flags);
893 smp_mb__after_clear_bit();
894 wake_up_bit(&nfsi->flags, NFS_INO_REVALIDATING);
895}
896
1da177e4
LT		 897int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
898{
899 struct inode *inode = dentry->d_inode;
55296809 900 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
1da177e4
LT		 901 int err;
902
903 if (__IS_FLG(inode, MS_NOATIME))
904 need_atime = 0;
905 else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode))
906 need_atime = 0;
907 /* We may force a getattr if the user cares about atime */
908 if (need_atime)
909 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
910 else
911 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
912 if (!err)
913 generic_fillattr(inode, stat);
914 return err;
915}
916
917struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred)
918{
919 struct nfs_open_context *ctx;
920
921 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
922 if (ctx != NULL) {
923 atomic_set(&ctx->count, 1);
924 ctx->dentry = dget(dentry);
925 ctx->cred = get_rpccred(cred);
926 ctx->state = NULL;
927 ctx->lockowner = current->files;
928 ctx->error = 0;
00a92642 929 ctx->dir_cookie = 0;
1da177e4
LT		 930 }
931 return ctx;
932}
933
934struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
935{
936 if (ctx != NULL)
937 atomic_inc(&ctx->count);
938 return ctx;
939}
940
941void put_nfs_open_context(struct nfs_open_context *ctx)
942{
943 if (atomic_dec_and_test(&ctx->count)) {
944 if (!list_empty(&ctx->list)) {
945 struct inode *inode = ctx->dentry->d_inode;
946 spin_lock(&inode->i_lock);
947 list_del(&ctx->list);
948 spin_unlock(&inode->i_lock);
949 }
950 if (ctx->state != NULL)
951 nfs4_close_state(ctx->state, ctx->mode);
952 if (ctx->cred != NULL)
953 put_rpccred(ctx->cred);
954 dput(ctx->dentry);
955 kfree(ctx);
956 }
957}
958
959/*
960 * Ensure that mmap has a recent RPC credential for use when writing out
961 * shared pages
962 */
963void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
964{
965 struct inode *inode = filp->f_dentry->d_inode;
966 struct nfs_inode *nfsi = NFS_I(inode);
967
968 filp->private_data = get_nfs_open_context(ctx);
969 spin_lock(&inode->i_lock);
970 list_add(&ctx->list, &nfsi->open_files);
971 spin_unlock(&inode->i_lock);
972}
973
974struct nfs_open_context *nfs_find_open_context(struct inode *inode, int mode)
975{
976 struct nfs_inode *nfsi = NFS_I(inode);
977 struct nfs_open_context *pos, *ctx = NULL;
978
979 spin_lock(&inode->i_lock);
980 list_for_each_entry(pos, &nfsi->open_files, list) {
981 if ((pos->mode & mode) == mode) {
982 ctx = get_nfs_open_context(pos);
983 break;
984 }
985 }
986 spin_unlock(&inode->i_lock);
987 return ctx;
988}
989
990void nfs_file_clear_open_context(struct file *filp)
991{
992 struct inode *inode = filp->f_dentry->d_inode;
993 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
994
995 if (ctx) {
996 filp->private_data = NULL;
997 spin_lock(&inode->i_lock);
998 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
999 spin_unlock(&inode->i_lock);
1000 put_nfs_open_context(ctx);
1001 }
1002}
1003
1004/*
1005 * These allocate and release file read/write context information.
1006 */
1007int nfs_open(struct inode *inode, struct file *filp)
1008{
1009 struct nfs_open_context *ctx;
1010 struct rpc_cred *cred;
1011
1012 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
1013 if (IS_ERR(cred))
1014 return PTR_ERR(cred);
1015 ctx = alloc_nfs_open_context(filp->f_dentry, cred);
1016 put_rpccred(cred);
1017 if (ctx == NULL)
1018 return -ENOMEM;
1019 ctx->mode = filp->f_mode;
1020 nfs_file_set_open_context(filp, ctx);
1021 put_nfs_open_context(ctx);
1022 if ((filp->f_mode & FMODE_WRITE) != 0)
1023 nfs_begin_data_update(inode);
1024 return 0;
1025}
1026
1027int nfs_release(struct inode *inode, struct file *filp)
1028{
1029 if ((filp->f_mode & FMODE_WRITE) != 0)
1030 nfs_end_data_update(inode);
1031 nfs_file_clear_open_context(filp);
1032 return 0;
1033}
1034
1035/*
1036 * This function is called whenever some part of NFS notices that
1037 * the cached attributes have to be refreshed.
1038 */
1039int
1040__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1041{
1042 int status = -ESTALE;
1043 struct nfs_fattr fattr;
1044 struct nfs_inode *nfsi = NFS_I(inode);
1045 unsigned long verifier;
55296809 1046 unsigned long cache_validity;
1da177e4
LT		 1047
1048 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1049 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1050
1051 lock_kernel();
1052 if (!inode || is_bad_inode(inode))
1053 goto out_nowait;
1054 if (NFS_STALE(inode))
1055 goto out_nowait;
1056
412d582e
CL		 1057 status = nfs_wait_on_inode(inode);
1058 if (status < 0)
1059 goto out;
1060 if (NFS_STALE(inode)) {
1061 status = -ESTALE;
1062 /* Do we trust the cached ESTALE? */
1063 if (NFS_ATTRTIMEO(inode) != 0) {
1064 if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME)) {
1065 /* no */
1066 } else
1067 goto out;
1068 }
1da177e4 1069 }
1da177e4
LT		 1070
1071 /* Protect against RPC races by saving the change attribute */
1072 verifier = nfs_save_change_attribute(inode);
1073 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1074 if (status != 0) {
1075 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1076 inode->i_sb->s_id,
1077 (long long)NFS_FILEID(inode), status);
1078 if (status == -ESTALE) {
1079 nfs_zap_caches(inode);
1080 if (!S_ISDIR(inode->i_mode))
412d582e 1081 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1da177e4
LT		 1082 }
1083 goto out;
1084 }
1085
1086 status = nfs_update_inode(inode, &fattr, verifier);
1087 if (status) {
1088 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1089 inode->i_sb->s_id,
1090 (long long)NFS_FILEID(inode), status);
1091 goto out;
1092 }
dc59250c 1093 spin_lock(&inode->i_lock);
55296809
CL		 1094 cache_validity = nfsi->cache_validity;
1095 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1096
1da177e4
LT		 1097 /*
1098 * We may need to keep the attributes marked as invalid if
1099 * we raced with nfs_end_attr_update().
1100 */
1101 if (verifier == nfsi->cache_change_attribute)
55296809 1102 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
dc59250c 1103 spin_unlock(&inode->i_lock);
55296809 1104
7d52e862 1105 nfs_revalidate_mapping(inode, inode->i_mapping);
55296809
CL		 1106
1107 if (cache_validity & NFS_INO_INVALID_ACL)
ada70d94 1108 nfs_zap_acl_cache(inode);
55296809 1109
1da177e4
LT		 1110 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1111 inode->i_sb->s_id,
1112 (long long)NFS_FILEID(inode));
1113
412d582e
CL		 1114 out:
1115 nfs_wake_up_inode(inode);
1116
1da177e4
LT		 1117 out_nowait:
1118 unlock_kernel();
1119 return status;
1120}
1121
1122int nfs_attribute_timeout(struct inode *inode)
1123{
1124 struct nfs_inode *nfsi = NFS_I(inode);
1125
1126 if (nfs_have_delegation(inode, FMODE_READ))
1127 return 0;
1128 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1129}
1130
1131/**
1132 * nfs_revalidate_inode - Revalidate the inode attributes
1133 * @server - pointer to nfs_server struct
1134 * @inode - pointer to inode struct
1135 *
1136 * Updates inode attribute information by retrieving the data from the server.
1137 */
1138int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1139{
55296809 1140 if (!(NFS_I(inode)->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1da177e4
LT		 1141 && !nfs_attribute_timeout(inode))
1142 return NFS_STALE(inode) ? -ESTALE : 0;
1143 return __nfs_revalidate_inode(server, inode);
1144}
1145
7d52e862
TM		 1146/**
1147 * nfs_revalidate_mapping - Revalidate the pagecache
1148 * @inode - pointer to host inode
1149 * @mapping - pointer to mapping
1150 */
1151void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
1152{
1153 struct nfs_inode *nfsi = NFS_I(inode);
1154
55296809 1155 if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
7d52e862
TM		 1156 if (S_ISREG(inode->i_mode)) {
1157 if (filemap_fdatawrite(mapping) == 0)
1158 filemap_fdatawait(mapping);
1159 nfs_wb_all(inode);
1160 }
1161 invalidate_inode_pages2(mapping);
dc59250c
CL		 1162
1163 spin_lock(&inode->i_lock);
55296809 1164 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
7d52e862
TM		 1165 if (S_ISDIR(inode->i_mode)) {
1166 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1167 /* This ensures we revalidate child dentries */
1168 nfsi->cache_change_attribute++;
1169 }
dc59250c
CL		 1170 spin_unlock(&inode->i_lock);
1171
7d52e862
TM		 1172 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1173 inode->i_sb->s_id,
1174 (long long)NFS_FILEID(inode));
1175 }
1176}
1177
1da177e4
LT		 1178/**
1179 * nfs_begin_data_update
1180 * @inode - pointer to inode
1181 * Declare that a set of operations will update file data on the server
1182 */
1183void nfs_begin_data_update(struct inode *inode)
1184{
1185 atomic_inc(&NFS_I(inode)->data_updates);
1186}
1187
1188/**
1189 * nfs_end_data_update
1190 * @inode - pointer to inode
1191 * Declare end of the operations that will update file data
1192 * This will mark the inode as immediately needing revalidation
1193 * of its attribute cache.
1194 */
1195void nfs_end_data_update(struct inode *inode)
1196{
1197 struct nfs_inode *nfsi = NFS_I(inode);
1198
1199 if (!nfs_have_delegation(inode, FMODE_READ)) {
1200 /* Mark the attribute cache for revalidation */
dc59250c 1201 spin_lock(&inode->i_lock);
55296809 1202 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4
LT		 1203 /* Directories and symlinks: invalidate page cache too */
1204 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
55296809 1205 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
dc59250c 1206 spin_unlock(&inode->i_lock);
1da177e4
LT		 1207 }
1208 nfsi->cache_change_attribute ++;
1209 atomic_dec(&nfsi->data_updates);
1210}
1211
1da177e4
LT		 1212/**
1213 * nfs_refresh_inode - verify consistency of the inode attribute cache
1214 * @inode - pointer to inode
1215 * @fattr - updated attributes
1216 *
1217 * Verifies the attribute cache. If we have just changed the attributes,
1218 * so that fattr carries weak cache consistency data, then it may
1219 * also update the ctime/mtime/change_attribute.
1220 */
1221int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1222{
1223 struct nfs_inode *nfsi = NFS_I(inode);
1224 loff_t cur_size, new_isize;
1225 int data_unstable;
1226
dc59250c
CL		 1227 spin_lock(&inode->i_lock);
1228
1da177e4
LT		 1229 /* Are we in the process of updating data on the server? */
1230 data_unstable = nfs_caches_unstable(inode);
1231
1232 if (fattr->valid & NFS_ATTR_FATTR_V4) {
1233 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1234 && nfsi->change_attr == fattr->pre_change_attr)
1235 nfsi->change_attr = fattr->change_attr;
fe51beec 1236 if (nfsi->change_attr != fattr->change_attr) {
55296809 1237 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1238 if (!data_unstable)
55296809 1239 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1240 }
1da177e4
LT		 1241 }
1242
dc59250c
CL		 1243 if ((fattr->valid & NFS_ATTR_FATTR) == 0) {
1244 spin_unlock(&inode->i_lock);
1da177e4 1245 return 0;
dc59250c 1246 }
1da177e4
LT		 1247
1248 /* Has the inode gone and changed behind our back? */
1249 if (nfsi->fileid != fattr->fileid
dc59250c
CL		 1250 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1251 spin_unlock(&inode->i_lock);
1da177e4 1252 return -EIO;
dc59250c 1253 }
1da177e4
LT		 1254
1255 cur_size = i_size_read(inode);
1256 new_isize = nfs_size_to_loff_t(fattr->size);
1257
1258 /* If we have atomic WCC data, we may update some attributes */
1259 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1260 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime))
1261 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1262 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime))
1263 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1264 }
1265
1266 /* Verify a few of the more important attributes */
fe51beec 1267 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
55296809 1268 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1269 if (!data_unstable)
55296809 1270 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec
TM		 1271 }
1272 if (cur_size != new_isize) {
55296809 1273 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1274 if (nfsi->npages == 0)
55296809 1275 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1276 }
1da177e4
LT		 1277
1278 /* Have any file permissions changed? */
1279 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1280 || inode->i_uid != fattr->uid
1281 || inode->i_gid != fattr->gid)
55296809 1282 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1da177e4
LT		 1283
1284 /* Has the link count changed? */
1285 if (inode->i_nlink != fattr->nlink)
55296809 1286 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4
LT		 1287
1288 if (!timespec_equal(&inode->i_atime, &fattr->atime))
55296809 1289 nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
1da177e4
LT		 1290
1291 nfsi->read_cache_jiffies = fattr->timestamp;
dc59250c 1292 spin_unlock(&inode->i_lock);
1da177e4
LT		 1293 return 0;
1294}
1295
1296/*
1297 * Many nfs protocol calls return the new file attributes after
1298 * an operation. Here we update the inode to reflect the state
1299 * of the server's inode.
1300 *
1301 * This is a bit tricky because we have to make sure all dirty pages
1302 * have been sent off to the server before calling invalidate_inode_pages.
1303 * To make sure no other process adds more write requests while we try
1304 * our best to flush them, we make them sleep during the attribute refresh.
1305 *
1306 * A very similar scenario holds for the dir cache.
1307 */
1308static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsigned long verifier)
1309{
1310 struct nfs_inode *nfsi = NFS_I(inode);
951a143b 1311 loff_t cur_isize, new_isize;
1da177e4 1312 unsigned int invalid = 0;
1da177e4
LT		 1313 int data_unstable;
1314
1315 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1316 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1317 atomic_read(&inode->i_count), fattr->valid);
1318
1319 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1320 return 0;
1321
1322 if (nfsi->fileid != fattr->fileid) {
1323 printk(KERN_ERR "%s: inode number mismatch\n"
1324 "expected (%s/0x%Lx), got (%s/0x%Lx)\n",
1325 __FUNCTION__,
1326 inode->i_sb->s_id, (long long)nfsi->fileid,
1327 inode->i_sb->s_id, (long long)fattr->fileid);
1328 goto out_err;
1329 }
1330
dc59250c
CL		 1331 spin_lock(&inode->i_lock);
1332
1da177e4
LT		 1333 /*
1334 * Make sure the inode's type hasn't changed.
1335 */
dc59250c
CL		 1336 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1337 spin_unlock(&inode->i_lock);
1da177e4 1338 goto out_changed;
dc59250c 1339 }
1da177e4
LT		 1340
1341 /*
1342 * Update the read time so we don't revalidate too often.
1343 */
1344 nfsi->read_cache_jiffies = fattr->timestamp;
1345
1346 /* Are we racing with known updates of the metadata on the server? */
b3c52da3
TM		 1347 data_unstable = ! (nfs_verify_change_attribute(inode, verifier) ||
1348 (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE));
1da177e4 1349
951a143b 1350 /* Check if our cached file size is stale */
1da177e4
LT		 1351 new_isize = nfs_size_to_loff_t(fattr->size);
1352 cur_isize = i_size_read(inode);
951a143b
TM		 1353 if (new_isize != cur_isize) {
1354 /* Do we perhaps have any outstanding writes? */
1355 if (nfsi->npages == 0) {
1356 /* No, but did we race with nfs_end_data_update()? */
1357 if (verifier == nfsi->cache_change_attribute) {
1da177e4 1358 inode->i_size = new_isize;
951a143b 1359 invalid |= NFS_INO_INVALID_DATA;
1da177e4 1360 }
951a143b
TM		 1361 invalid |= NFS_INO_INVALID_ATTR;
1362 } else if (new_isize > cur_isize) {
1da177e4
LT		 1363 inode->i_size = new_isize;
1364 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1365 }
951a143b
TM		 1366 dprintk("NFS: isize change on server for file %s/%ld\n",
1367 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1368 }
1369
951a143b 1370 /* Check if the mtime agrees */
1da177e4
LT		 1371 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1372 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
951a143b
TM		 1373 dprintk("NFS: mtime change on server for file %s/%ld\n",
1374 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1375 if (!data_unstable)
1376 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1377 }
1378
1379 if ((fattr->valid & NFS_ATTR_FATTR_V4)
1380 && nfsi->change_attr != fattr->change_attr) {
951a143b 1381 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1da177e4 1382 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT		 1383 nfsi->change_attr = fattr->change_attr;
1384 if (!data_unstable)
ada70d94 1385 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1386 }
1387
ada70d94
TM		 1388 /* If ctime has changed we should definitely clear access+acl caches */
1389 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
1390 if (!data_unstable)
1391 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1392 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1393 }
1da177e4
LT		 1394 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1395
1396 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1397 inode->i_uid != fattr->uid ||
1398 inode->i_gid != fattr->gid)
ada70d94 1399 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1400
1401 inode->i_mode = fattr->mode;
1402 inode->i_nlink = fattr->nlink;
1403 inode->i_uid = fattr->uid;
1404 inode->i_gid = fattr->gid;
1405
1406 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1407 /*
1408 * report the blocks in 512byte units
1409 */
1410 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1411 inode->i_blksize = inode->i_sb->s_blocksize;
1412 } else {
1413 inode->i_blocks = fattr->du.nfs2.blocks;
1414 inode->i_blksize = fattr->du.nfs2.blocksize;
1415 }
1416
1417 /* Update attrtimeo value if we're out of the unstable period */
1418 if (invalid & NFS_INO_INVALID_ATTR) {
1419 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1420 nfsi->attrtimeo_timestamp = jiffies;
1421 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1422 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1423 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1424 nfsi->attrtimeo_timestamp = jiffies;
1425 }
1426 /* Don't invalidate the data if we were to blame */
1427 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1428 || S_ISLNK(inode->i_mode)))
1429 invalid &= ~NFS_INO_INVALID_DATA;
1430 if (!nfs_have_delegation(inode, FMODE_READ))
55296809 1431 nfsi->cache_validity |= invalid;
1da177e4 1432
dc59250c 1433 spin_unlock(&inode->i_lock);
1da177e4
LT		 1434 return 0;
1435 out_changed:
1436 /*
1437 * Big trouble! The inode has become a different object.
1438 */
1439#ifdef NFS_PARANOIA
1440 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1441 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1442#endif
1443 /*
1444 * No need to worry about unhashing the dentry, as the
1445 * lookup validation will know that the inode is bad.
1446 * (But we fall through to invalidate the caches.)
1447 */
1448 nfs_invalidate_inode(inode);
1449 out_err:
412d582e 1450 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1da177e4
LT		 1451 return -ESTALE;
1452}
1453
1454/*
1455 * File system information
1456 */
1457
1458static int nfs_set_super(struct super_block *s, void *data)
1459{
1460 s->s_fs_info = data;
1461 return set_anon_super(s, data);
1462}
1463
1464static int nfs_compare_super(struct super_block *sb, void *data)
1465{
1466 struct nfs_server *server = data;
1467 struct nfs_server *old = NFS_SB(sb);
1468
1469 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1470 return 0;
1471 if (old->addr.sin_port != server->addr.sin_port)
1472 return 0;
1473 return !nfs_compare_fh(&old->fh, &server->fh);
1474}
1475
1476static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1477 int flags, const char *dev_name, void *raw_data)
1478{
1479 int error;
9085bbcb 1480 struct nfs_server *server = NULL;
1da177e4
LT		 1481 struct super_block *s;
1482 struct nfs_fh *root;
1483 struct nfs_mount_data *data = raw_data;
1484
9085bbcb
TM		 1485 s = ERR_PTR(-EINVAL);
1486 if (data == NULL) {
1487 dprintk("%s: missing data argument\n", __FUNCTION__);
1488 goto out_err;
1489 }
1490 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1491 dprintk("%s: bad mount version\n", __FUNCTION__);
1492 goto out_err;
1da177e4 1493 }
9085bbcb
TM		 1494 switch (data->version) {
1495 case 1:
1496 data->namlen = 0;
1497 case 2:
1498 data->bsize = 0;
1499 case 3:
1500 if (data->flags & NFS_MOUNT_VER3) {
1501 dprintk("%s: mount structure version %d does not support NFSv3\n",
1502 __FUNCTION__,
1503 data->version);
1504 goto out_err;
1505 }
1506 data->root.size = NFS2_FHSIZE;
1507 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1508 case 4:
1509 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1510 dprintk("%s: mount structure version %d does not support strong security\n",
1511 __FUNCTION__,
1512 data->version);
1513 goto out_err;
1514 }
1515 case 5:
1516 memset(data->context, 0, sizeof(data->context));
1517 }
1518#ifndef CONFIG_NFS_V3
1519 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1520 s = ERR_PTR(-EPROTONOSUPPORT);
1521 if (data->flags & NFS_MOUNT_VER3) {
1522 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1523 goto out_err;
1524 }
1525#endif /* CONFIG_NFS_V3 */
1da177e4 1526
9085bbcb 1527 s = ERR_PTR(-ENOMEM);
1da177e4
LT		 1528 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1529 if (!server)
9085bbcb 1530 goto out_err;
1da177e4
LT		 1531 memset(server, 0, sizeof(struct nfs_server));
1532 /* Zero out the NFS state stuff */
1533 init_nfsv4_state(server);
b7fa0554 1534 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1535
1da177e4
LT		 1536 root = &server->fh;
1537 if (data->flags & NFS_MOUNT_VER3)
1538 root->size = data->root.size;
1539 else
1540 root->size = NFS2_FHSIZE;
9085bbcb 1541 s = ERR_PTR(-EINVAL);
1da177e4 1542 if (root->size > sizeof(root->data)) {
9085bbcb
TM		 1543 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1544 goto out_err;
1da177e4
LT		 1545 }
1546 memcpy(root->data, data->root.data, root->size);
1547
1548 /* We now require that the mount process passes the remote address */
1549 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1550 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1551 dprintk("%s: mount program didn't pass remote address!\n",
1552 __FUNCTION__);
1553 goto out_err;
1da177e4
LT		 1554 }
1555
9085bbcb
TM		 1556 /* Fire up rpciod if not yet running */
1557 s = ERR_PTR(rpciod_up());
1558 if (IS_ERR(s)) {
1559 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1560 __FUNCTION__, PTR_ERR(s));
1561 goto out_err;
1da177e4
LT		 1562 }
1563
9085bbcb
TM		 1564 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1565 if (IS_ERR(s) || s->s_root)
1566 goto out_rpciod_down;
1da177e4 1567
9085bbcb 1568 s->s_flags = flags;
1da177e4
LT		 1569
1570 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1571 if (error) {
1572 up_write(&s->s_umount);
1573 deactivate_super(s);
1574 return ERR_PTR(error);
1575 }
1576 s->s_flags |= MS_ACTIVE;
1577 return s;
9085bbcb
TM		 1578out_rpciod_down:
1579 rpciod_down();
1580out_err:
1581 kfree(server);
1582 return s;
1da177e4
LT		 1583}
1584
1585static void nfs_kill_super(struct super_block *s)
1586{
1587 struct nfs_server *server = NFS_SB(s);
1588
1589 kill_anon_super(s);
1590
6a19275a 1591 if (!IS_ERR(server->client))
1da177e4 1592 rpc_shutdown_client(server->client);
6a19275a 1593 if (!IS_ERR(server->client_sys))
1da177e4 1594 rpc_shutdown_client(server->client_sys);
b7fa0554
AG		 1595 if (!IS_ERR(server->client_acl))
1596 rpc_shutdown_client(server->client_acl);
1da177e4
LT		 1597
1598 if (!(server->flags & NFS_MOUNT_NONLM))
1599 lockd_down(); /* release rpc.lockd */
1600
1601 rpciod_down(); /* release rpciod */
1602
1603 if (server->hostname != NULL)
1604 kfree(server->hostname);
1605 kfree(server);
1606}
1607
1608static struct file_system_type nfs_fs_type = {
1609 .owner = THIS_MODULE,
1610 .name = "nfs",
1611 .get_sb = nfs_get_sb,
1612 .kill_sb = nfs_kill_super,
1613 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1614};
1615
1616#ifdef CONFIG_NFS_V4
1617
1618static void nfs4_clear_inode(struct inode *);
1619
1620
1621static struct super_operations nfs4_sops = {
1622 .alloc_inode = nfs_alloc_inode,
1623 .destroy_inode = nfs_destroy_inode,
1624 .write_inode = nfs_write_inode,
1625 .delete_inode = nfs_delete_inode,
1626 .statfs = nfs_statfs,
1627 .clear_inode = nfs4_clear_inode,
1628 .umount_begin = nfs_umount_begin,
1629 .show_options = nfs_show_options,
1630};
1631
1632/*
1633 * Clean out any remaining NFSv4 state that might be left over due
1634 * to open() calls that passed nfs_atomic_lookup, but failed to call
1635 * nfs_open().
1636 */
1637static void nfs4_clear_inode(struct inode *inode)
1638{
1639 struct nfs_inode *nfsi = NFS_I(inode);
1640
1641 /* If we are holding a delegation, return it! */
1642 if (nfsi->delegation != NULL)
1643 nfs_inode_return_delegation(inode);
1644 /* First call standard NFS clear_inode() code */
1645 nfs_clear_inode(inode);
1646 /* Now clear out any remaining state */
1647 while (!list_empty(&nfsi->open_states)) {
1648 struct nfs4_state *state;
1649
1650 state = list_entry(nfsi->open_states.next,
1651 struct nfs4_state,
1652 inode_states);
1653 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1654 __FUNCTION__,
1655 inode->i_sb->s_id,
1656 (long long)NFS_FILEID(inode),
1657 state);
1658 BUG_ON(atomic_read(&state->count) != 1);
1659 nfs4_close_state(state, state->state);
1660 }
1661}
1662
1663
1664static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1665{
1666 struct nfs_server *server;
1667 struct nfs4_client *clp = NULL;
1668 struct rpc_xprt *xprt = NULL;
1669 struct rpc_clnt *clnt = NULL;
1670 struct rpc_timeout timeparms;
1671 rpc_authflavor_t authflavour;
1672 int proto, err = -EIO;
1673
1674 sb->s_blocksize_bits = 0;
1675 sb->s_blocksize = 0;
1676 server = NFS_SB(sb);
1677 if (data->rsize != 0)
1678 server->rsize = nfs_block_size(data->rsize, NULL);
1679 if (data->wsize != 0)
1680 server->wsize = nfs_block_size(data->wsize, NULL);
1681 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1682 server->caps = NFS_CAP_ATOMIC_OPEN;
1683
1684 server->acregmin = data->acregmin*HZ;
1685 server->acregmax = data->acregmax*HZ;
1686 server->acdirmin = data->acdirmin*HZ;
1687 server->acdirmax = data->acdirmax*HZ;
1688
1689 server->rpc_ops = &nfs_v4_clientops;
1690 /* Initialize timeout values */
1691
1692 timeparms.to_initval = data->timeo * HZ / 10;
1693 timeparms.to_retries = data->retrans;
1694 timeparms.to_exponential = 1;
1695 if (!timeparms.to_retries)
1696 timeparms.to_retries = 5;
1697
1698 proto = data->proto;
1699 /* Which IP protocol do we use? */
1700 switch (proto) {
1701 case IPPROTO_TCP:
1702 timeparms.to_maxval = RPC_MAX_TCP_TIMEOUT;
1703 if (!timeparms.to_initval)
1704 timeparms.to_initval = 600 * HZ / 10;
1705 break;
1706 case IPPROTO_UDP:
1707 timeparms.to_maxval = RPC_MAX_UDP_TIMEOUT;
1708 if (!timeparms.to_initval)
1709 timeparms.to_initval = 11 * HZ / 10;
1710 break;
1711 default:
1712 return -EINVAL;
1713 }
1714
1715 clp = nfs4_get_client(&server->addr.sin_addr);
1716 if (!clp) {
9085bbcb 1717 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1da177e4
LT		 1718 return -EIO;
1719 }
1720
1721 /* Now create transport and client */
1722 authflavour = RPC_AUTH_UNIX;
1723 if (data->auth_flavourlen != 0) {
9085bbcb
TM		 1724 if (data->auth_flavourlen != 1) {
1725 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1726 __FUNCTION__, data->auth_flavourlen);
1727 err = -EINVAL;
1728 goto out_fail;
1729 }
1da177e4
LT		 1730 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1731 err = -EFAULT;
1732 goto out_fail;
1733 }
1734 }
1735
1736 down_write(&clp->cl_sem);
6a19275a 1737 if (IS_ERR(clp->cl_rpcclient)) {
1da177e4
LT		 1738 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1739 if (IS_ERR(xprt)) {
1740 up_write(&clp->cl_sem);
1da177e4 1741 err = PTR_ERR(xprt);
9085bbcb
TM		 1742 dprintk("%s: cannot create RPC transport. Error = %d\n",
1743 __FUNCTION__, err);
1da177e4
LT		 1744 goto out_fail;
1745 }
1746 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1747 server->rpc_ops->version, authflavour);
1748 if (IS_ERR(clnt)) {
1749 up_write(&clp->cl_sem);
1da177e4 1750 err = PTR_ERR(clnt);
9085bbcb
TM		 1751 dprintk("%s: cannot create RPC client. Error = %d\n",
1752 __FUNCTION__, err);
1da177e4
LT		 1753 goto out_fail;
1754 }
1755 clnt->cl_intr = 1;
1756 clnt->cl_softrtry = 1;
1757 clnt->cl_chatty = 1;
1758 clp->cl_rpcclient = clnt;
1759 clp->cl_cred = rpcauth_lookupcred(clnt->cl_auth, 0);
1760 if (IS_ERR(clp->cl_cred)) {
1761 up_write(&clp->cl_sem);
1762 err = PTR_ERR(clp->cl_cred);
1763 clp->cl_cred = NULL;
1764 goto out_fail;
1765 }
1766 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1767 nfs_idmap_new(clp);
1768 }
1769 if (list_empty(&clp->cl_superblocks)) {
1770 err = nfs4_init_client(clp);
1771 if (err != 0) {
1772 up_write(&clp->cl_sem);
1773 goto out_fail;
1774 }
1775 }
1776 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1777 clnt = rpc_clone_client(clp->cl_rpcclient);
1778 if (!IS_ERR(clnt))
1779 server->nfs4_state = clp;
1780 up_write(&clp->cl_sem);
1781 clp = NULL;
1782
1783 if (IS_ERR(clnt)) {
9085bbcb
TM		 1784 err = PTR_ERR(clnt);
1785 dprintk("%s: cannot create RPC client. Error = %d\n",
1786 __FUNCTION__, err);
1787 return err;
1da177e4
LT		 1788 }
1789
1790 server->client = clnt;
1791
1792 if (server->nfs4_state->cl_idmap == NULL) {
9085bbcb 1793 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1da177e4
LT		 1794 return -ENOMEM;
1795 }
1796
1797 if (clnt->cl_auth->au_flavor != authflavour) {
6a19275a
BF		 1798 struct rpc_auth *auth;
1799
1800 auth = rpcauth_create(authflavour, clnt);
1801 if (IS_ERR(auth)) {
9085bbcb 1802 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
6a19275a 1803 return PTR_ERR(auth);
1da177e4
LT		 1804 }
1805 }
1806
1807 sb->s_time_gran = 1;
1808
1809 sb->s_op = &nfs4_sops;
1810 err = nfs_sb_init(sb, authflavour);
1811 if (err == 0)
1812 return 0;
1813out_fail:
1814 if (clp)
1815 nfs4_put_client(clp);
1816 return err;
1817}
1818
1819static int nfs4_compare_super(struct super_block *sb, void *data)
1820{
1821 struct nfs_server *server = data;
1822 struct nfs_server *old = NFS_SB(sb);
1823
1824 if (strcmp(server->hostname, old->hostname) != 0)
1825 return 0;
1826 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1827 return 0;
1828 return 1;
1829}
1830
1831static void *
1832nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1833{
1834 void *p = NULL;
1835
1836 if (!src->len)
1837 return ERR_PTR(-EINVAL);
1838 if (src->len < maxlen)
1839 maxlen = src->len;
1840 if (dst == NULL) {
1841 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1842 if (p == NULL)
1843 return ERR_PTR(-ENOMEM);
1844 }
1845 if (copy_from_user(dst, src->data, maxlen)) {
1846 if (p != NULL)
1847 kfree(p);
1848 return ERR_PTR(-EFAULT);
1849 }
1850 dst[maxlen] = '\0';
1851 return dst;
1852}
1853
1854static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1855 int flags, const char *dev_name, void *raw_data)
1856{
1857 int error;
1858 struct nfs_server *server;
1859 struct super_block *s;
1860 struct nfs4_mount_data *data = raw_data;
1861 void *p;
1862
9085bbcb
TM		 1863 if (data == NULL) {
1864 dprintk("%s: missing data argument\n", __FUNCTION__);
1865 return ERR_PTR(-EINVAL);
1866 }
1867 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1868 dprintk("%s: bad mount version\n", __FUNCTION__);
1da177e4
LT		 1869 return ERR_PTR(-EINVAL);
1870 }
1871
1872 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1873 if (!server)
1874 return ERR_PTR(-ENOMEM);
1875 memset(server, 0, sizeof(struct nfs_server));
1876 /* Zero out the NFS state stuff */
1877 init_nfsv4_state(server);
b7fa0554 1878 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1879
1da177e4
LT		 1880 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1881 if (IS_ERR(p))
1882 goto out_err;
1883 server->hostname = p;
1884
1885 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1886 if (IS_ERR(p))
1887 goto out_err;
1888 server->mnt_path = p;
1889
1890 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1891 sizeof(server->ip_addr) - 1);
1892 if (IS_ERR(p))
1893 goto out_err;
1894
1895 /* We now require that the mount process passes the remote address */
1896 if (data->host_addrlen != sizeof(server->addr)) {
1897 s = ERR_PTR(-EINVAL);
1898 goto out_free;
1899 }
1900 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1901 s = ERR_PTR(-EFAULT);
1902 goto out_free;
1903 }
1904 if (server->addr.sin_family != AF_INET ||
1905 server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1906 dprintk("%s: mount program didn't pass remote IP address!\n",
1907 __FUNCTION__);
1da177e4
LT		 1908 s = ERR_PTR(-EINVAL);
1909 goto out_free;
1910 }
1911
9085bbcb
TM		 1912 /* Fire up rpciod if not yet running */
1913 s = ERR_PTR(rpciod_up());
1914 if (IS_ERR(s)) {
1915 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1916 __FUNCTION__, PTR_ERR(s));
1917 goto out_free;
1918 }
1919
1da177e4
LT		 1920 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1921
1922 if (IS_ERR(s) || s->s_root)
1923 goto out_free;
1924
1925 s->s_flags = flags;
1926
1da177e4
LT		 1927 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1928 if (error) {
1929 up_write(&s->s_umount);
1930 deactivate_super(s);
1931 return ERR_PTR(error);
1932 }
1933 s->s_flags |= MS_ACTIVE;
1934 return s;
1935out_err:
1936 s = (struct super_block *)p;
1937out_free:
1938 if (server->mnt_path)
1939 kfree(server->mnt_path);
1940 if (server->hostname)
1941 kfree(server->hostname);
1942 kfree(server);
1943 return s;
1944}
1945
1946static void nfs4_kill_super(struct super_block *sb)
1947{
1948 struct nfs_server *server = NFS_SB(sb);
1949
1950 nfs_return_all_delegations(sb);
1951 kill_anon_super(sb);
1952
1953 nfs4_renewd_prepare_shutdown(server);
1954
1955 if (server->client != NULL && !IS_ERR(server->client))
1956 rpc_shutdown_client(server->client);
1957 rpciod_down(); /* release rpciod */
1958
1959 destroy_nfsv4_state(server);
1960
1961 if (server->hostname != NULL)
1962 kfree(server->hostname);
1963 kfree(server);
1964}
1965
1966static struct file_system_type nfs4_fs_type = {
1967 .owner = THIS_MODULE,
1968 .name = "nfs4",
1969 .get_sb = nfs4_get_sb,
1970 .kill_sb = nfs4_kill_super,
1971 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1972};
1973
1974#define nfs4_init_once(nfsi) \
1975 do { \
1976 INIT_LIST_HEAD(&(nfsi)->open_states); \
1977 nfsi->delegation = NULL; \
1978 nfsi->delegation_state = 0; \
1979 init_rwsem(&nfsi->rwsem); \
1980 } while(0)
1981#define register_nfs4fs() register_filesystem(&nfs4_fs_type)
1982#define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type)
1983#else
1984#define nfs4_init_once(nfsi) \
1985 do { } while (0)
1986#define register_nfs4fs() (0)
1987#define unregister_nfs4fs()
1988#endif
1989
1990extern int nfs_init_nfspagecache(void);
1991extern void nfs_destroy_nfspagecache(void);
1992extern int nfs_init_readpagecache(void);
1993extern void nfs_destroy_readpagecache(void);
1994extern int nfs_init_writepagecache(void);
1995extern void nfs_destroy_writepagecache(void);
1996#ifdef CONFIG_NFS_DIRECTIO
1997extern int nfs_init_directcache(void);
1998extern void nfs_destroy_directcache(void);
1999#endif
2000
2001static kmem_cache_t * nfs_inode_cachep;
2002
2003static struct inode *nfs_alloc_inode(struct super_block *sb)
2004{
2005 struct nfs_inode *nfsi;
2006 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
2007 if (!nfsi)
2008 return NULL;
55296809
CL		 2009 nfsi->flags = 0UL;
2010 nfsi->cache_validity = 0UL;
458818ed
TM		 2011#ifdef CONFIG_NFS_V3_ACL
2012 nfsi->acl_access = ERR_PTR(-EAGAIN);
2013 nfsi->acl_default = ERR_PTR(-EAGAIN);
2014#endif
e50a1c2e
BF		 2015#ifdef CONFIG_NFS_V4
2016 nfsi->nfs4_acl = NULL;
2017#endif /* CONFIG_NFS_V4 */
1da177e4
LT		 2018 return &nfsi->vfs_inode;
2019}
2020
2021static void nfs_destroy_inode(struct inode *inode)
2022{
2023 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
2024}
2025
2026static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
2027{
2028 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
2029
2030 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
2031 SLAB_CTOR_CONSTRUCTOR) {
2032 inode_init_once(&nfsi->vfs_inode);
2033 spin_lock_init(&nfsi->req_lock);
2034 INIT_LIST_HEAD(&nfsi->dirty);
2035 INIT_LIST_HEAD(&nfsi->commit);
2036 INIT_LIST_HEAD(&nfsi->open_files);
2037 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
2038 atomic_set(&nfsi->data_updates, 0);
2039 nfsi->ndirty = 0;
2040 nfsi->ncommit = 0;
2041 nfsi->npages = 0;
1da177e4
LT		 2042 nfs4_init_once(nfsi);
2043 }
2044}
2045
75c96f85 2046static int nfs_init_inodecache(void)
1da177e4
LT		 2047{
2048 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2049 sizeof(struct nfs_inode),
2050 0, SLAB_RECLAIM_ACCOUNT,
2051 init_once, NULL);
2052 if (nfs_inode_cachep == NULL)
2053 return -ENOMEM;
2054
2055 return 0;
2056}
2057
75c96f85 2058static void nfs_destroy_inodecache(void)
1da177e4
LT		 2059{
2060 if (kmem_cache_destroy(nfs_inode_cachep))
2061 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2062}
2063
2064/*
2065 * Initialize NFS
2066 */
2067static int __init init_nfs_fs(void)
2068{
2069 int err;
2070
2071 err = nfs_init_nfspagecache();
2072 if (err)
2073 goto out4;
2074
2075 err = nfs_init_inodecache();
2076 if (err)
2077 goto out3;
2078
2079 err = nfs_init_readpagecache();
2080 if (err)
2081 goto out2;
2082
2083 err = nfs_init_writepagecache();
2084 if (err)
2085 goto out1;
2086
2087#ifdef CONFIG_NFS_DIRECTIO
2088 err = nfs_init_directcache();
2089 if (err)
2090 goto out0;
2091#endif
2092
2093#ifdef CONFIG_PROC_FS
2094 rpc_proc_register(&nfs_rpcstat);
2095#endif
2096 err = register_filesystem(&nfs_fs_type);
2097 if (err)
2098 goto out;
2099 if ((err = register_nfs4fs()) != 0)
2100 goto out;
2101 return 0;
2102out:
2103#ifdef CONFIG_PROC_FS
2104 rpc_proc_unregister("nfs");
2105#endif
2106 nfs_destroy_writepagecache();
2107#ifdef CONFIG_NFS_DIRECTIO
2108out0:
2109 nfs_destroy_directcache();
2110#endif
2111out1:
2112 nfs_destroy_readpagecache();
2113out2:
2114 nfs_destroy_inodecache();
2115out3:
2116 nfs_destroy_nfspagecache();
2117out4:
2118 return err;
2119}
2120
2121static void __exit exit_nfs_fs(void)
2122{
2123#ifdef CONFIG_NFS_DIRECTIO
2124 nfs_destroy_directcache();
2125#endif
2126 nfs_destroy_writepagecache();
2127 nfs_destroy_readpagecache();
2128 nfs_destroy_inodecache();
2129 nfs_destroy_nfspagecache();
2130#ifdef CONFIG_PROC_FS
2131 rpc_proc_unregister("nfs");
2132#endif
2133 unregister_filesystem(&nfs_fs_type);
2134 unregister_nfs4fs();
2135}
2136
2137/* Not quite true; I just maintain it */
2138MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2139MODULE_LICENSE("GPL");
2140
2141module_init(init_nfs_fs)
2142module_exit(exit_nfs_fs)
Linux 6.x block layer and io_uring tree(s)